Prediction of response to adrenalectomy in the treatment of advanced breast cancer.

Abstract

A collaborative study was done that was aimed at the development of a prognostic index for the assignment of patients with metastatic breast carcinoma to bilateral adrenalectomy for palliation. A total of 121 patients treated by adrenalectomy was analyzed; 28 remissions occurred for an overall remission rate of 23%. When used in its original form, Bulbrook's discriminant was not useful as a predictive index for the patients. Only 21 of 88 patients (24%) with positive discriminants had a remission, whereas 7 of 33 patients (21%) with negative discriminants had a remission. Replacing etiocholanolone with androsterone improved the performance of the discriminant, inasmuch as 21 of 73 patients (29%) with positive discriminants had a remission and 7 of 48 patients (15%) with negative discriminants had a remission; however, the association of the discriminant with remission was not statistically significant. Our approach to the development of a new predictive index was based on the logistic regression model as representing the relationship between concomitant information and the probability of remission. Three steroid quantities were found to have an apparent relationship with remission: androsterone plus etlocholanolone, total 17-ketogenic steroids, and 3α, 17α, 21-trihydroxy-5 β-pregnane-11, 20-dione/3 α, 17 α, 21-trihydroxy-5 α-pregnane-11,20-dione. However, the relationships of the three steroid quantities to remission were not highly statistically significant, which raised-doubts as to whether the results obtained here would be replicated with another set of data. The comparison of our findings with those of a recent study on the use of estrogen receptor determinations in predicting response to endocrine therapy suggested that, for this purpose, such determinations were superior to the measurement of urinary steroid metabolites. However, because of the selection that occurs in the identification of patients for adrenalectomy and the variation in the history of previous treatments, results of the study of response to endocrine therapy and the development of predictors may have little general application because the assumption that a study group of patients is a sample from a definable population may have little validity.